This talk will present Elasticsearch as a capable vector database. It will explain the conceptual and technical differences between semantic and lexical search as well as the concept of embeddings and semantic similarity. Using data from Czech Wikipedia, the talk will focus on code-based demonstration of generating embeddings, indexing data and performing searches, with a particular focus on performance, as well as evaluating the quality of search results.
Timestamps:
00:00 Introduction
02:00 What is Semantic Search?
05:43 What are Text Embeddings?
09:50 What is Semantic Similarity?
11:52 Generating Embeddings
17:05 How to Work with Vectors in Elasticsearch
19:26 Step 1: Pre-Process the Data
28:35 Step 2: Generate Embeddings
32:54 Step 3: Index Embeddings
40:00 Step 4: Search the Data
52:03 Step 5: Evaluate Search Quality
Speaker: Karel Minařík
Additional resource: github.com/karmi/talks
Make sure to join your local Elastic User Group to stay up-to-date on upcoming meetups: community.elastic.co/
Questions? Check out discuss.elastic.co/
Connect with the Elastic community through Slack: ela.st/slack
#semanticsearch #elasticsearch #vectordatabase #lexicalsearch #search #semantic #embeddings #techcommunity #techtalk
Негізгі бет Фильм және анимация Semantic Search With Elasticsearch
Пікірлер: 2